Supervized segmentation with graph-structured deep metric learning

Loic Landrieu (1) , Mohamed Boussaha (1)

1 LaSTIG - Laboratoire des Sciences et Technologies de l'Information Géographique

We present a fully-supervized method for learning to segment data structured by an adjacency graph. We introduce the graph-structured contrastive loss, a loss function structured by a ground truth segmentation. It promotes learning vertex embed-dings which are homogeneous within desired segments , and have high contrast at their interface. Thus, computing a piecewise-constant approximation of such embeddings produces a graph-partition close to the objective segmentation. This loss is fully backpropagable, which allows us to learn vertex embeddings with deep learning algorithms. We evaluate our methods on a 3D point cloud oversegmentation task, defining a new state-of-the-art by a large margin. These results are based on the published work of Landrieu & Bous-saha (2019).

Domaines: Machine Learning [stat.ML]; Réseau de neurones [cs.NE]

Loic Landrieu, Mohamed Boussaha. Supervized segmentation with graph-structured deep metric learning. ICML Workshop on Learning and Reasoning with Graph-Structured Representations, Jun 2019, Long Beach (CA), United States. ⟨hal-03016114⟩